INTELEC 2002 Paper 32.

1 1

Life Extension Through Charge Equalization of

Lead-Acid Batteries
Philip T. Krein, Fellow, IEEE, Robert Balog, Student Member, IEEE

that “Pack imbalance is perhaps the most serious issue with

Abstract—Charge equalization is an important part of the large series string packs.”
charge process for series-connected battery cells. This paper Although much of the research with respect to equalization
reviews battery behavior and performance related to the has considered long series strings (100 or more cells in series),
equalization problem, in the context of vavle-regulated lead-acid
batteries. As established in prior work, equalization precision on
problems of equalization extend down to short strings of just a
the order of 10 mV/cell is required for a successful process. few cells. In [3], a conventional block of six lead-acid cells
Equalization processes that can achieve this precision do indeed was tested with a voltage-limited charge process, in which the
extend the life of battery packs. Active equalization speeds the maximum charge potential is held below 2.35 V for any
process and supports exchange of a single failed cell or individual cell. When a 12-hour charge was used in this case,
monoblock Passive equalization (conventional overcharge) is too performance degraded from cycle to cycle. Only when the
slow in most contexts for strings longer than about 12 cells.
Active equalization provides clear life advantages over other
charge process was extended to 16 hr was there enough time
approaches. The results are consistent with the hypothesis that a for equalization to bring about recovery. Even then,
properly designed active equalization process can provide approximately one week would have been needed to fully
battery cycle life for a series string that matches the cycle restore charge balance. The equalization process represented
operating results for an individual cell. by this charge strategy reflects a “passive equalization”
approach: ensure that strong cells are subjected to overcharge
Index Terms— Battery equalization, charge equalization, (limited by voltage in [3]) until weak cells regain full charge.
battery management, charge balancing
The process is slow even for familiar 12 V monoblocks.
Several key questions must be considered with respect to
I. INTRODUCTION equalization:
1. Many equalization processes match cell voltages. The
Batteries are nearly always used in series combinations of real need is to match cell state-of-charge (SOC). Is voltage an
multiple cells. When a series string of cells is charged as a accurate surrogate for SOC?
group, a single current is imposed on all the cells. However, if 2. How accurately must the cells be matched to make the
the voltages begin to differ, the result is a charge imbalance process useful?
that can lead ultimately to battery failure. In any useful series 3. What are the benefits of equalization?
battery charge process, some type of “charge balancing” or 4. Will equalization extend battery life and reduce costs?
“equalization” must take place to restore balance or at least These questions are addressed in this paper, primarily in the
prevent it from growing. context of modern valve-regulated lead-acid (VRLA)
The need for equalization is well established. In most batteries.
conventional battery charging practice, equalization is
addressed either by driving the charge to a sufficient potential II. VOLTAGE BALANCING
to assure some degree of overcharge for all cells, or with a
separate higher-voltage charging step intended to reach the An equalization process is intended to match the SOC
weakest cells. In recent years, electric and hybrid vehicle among cells in a string. Since the charge current matches in a
applications, which tend to use very long series strings and series string, the implication is that voltage should match.
which push battery performance to extremes, have brought However, the relationship between voltage and SOC is not
charge equalization into wider view. In [1], results with lead- necessarily trivial. Fig. 1 shows a graph (from [4]) of cell
acid battery packs in the General Motors EV1 production voltage vs. SOC for a lead-acid cell. The relationship is based
electric car are presented. The paper focuses most of its on measuring open-circuit voltage after about 30 min of rest.
attention on the charge equalization problem. In [2], a team It shows a simple linear relationship. The total voltage change
from Optima Battery (now part of Johnson Controls) reports over a 0 to 100% SOC change is about 0.2 V, corresponding
to 2 mV for every 1% change in SOC. Two key inferences
The authors are with the Grainger Center for Electric Machinery and can be made from this graph:
Electromechanics, University of Illinois, Department of Electrical and
Computer Engineering, 1406 W. Green St, Urbana, IL 61801 USA (e-mail 1. For lead-acid cells in steady state, voltage provides
machines@ece.uiuc.edu). useful information about SOC.
INTELEC 2002 Paper 32.1 2

2. Cell-to-cell voltage matching on the order of 10 mV 5

corresponds (in steady state) to SOC match on the order of
4.8
5%.
4.6

AH Capacity at 3 h Rate
2.15
4.4
Test pack 1
4.2 Test pack 2
2.1
Open Circuit Voltage(V)

2.05 3.8

3.6

2 3.4

3.2
1.95
3
0 5 10 15 20 25 30 35
Cycle Number
1.9
0 10 20 30 40 50 60 70 80 90 100 Fig. 3. Cell capacity vs. cycle number for comparison test ([5]).
State of Charge(%)
Fig. 1. VRLA open-circuit voltage vs. state of charge (from [4]). Based on the steady-state voltage behavior in Fig. 1 and the
dynamic analysis in [5], we conclude that voltage matching is
It is a challenge to interpret this behavior under dynamic an effective way to match SOC for lead-acid batteries. This is
conditions. A reasonable assumption is that low currents will true provided the currents are limited (the equalization current
reflect similar behavior, although high currents will distort the used in [5] is about 1% of the battery’s nominal C rate).
voltage behavior and could lead to erroneous results. The conclusion: To be useful, an equalization process
Experimental life tests have been conducted to evaluate the should hold the cell-to-cell differences to about 10 mV or less.
voltage-matching behavior of VRLA batteries under dynamic This is consistent with other results from the literature. In [8],
conditions. Figs. 2 and 3 (from [5]) illustrate the results. The for example, a large set of batteries is monitored cell-by-cell
figures present test results for two 12-cell VRLA strings tested for condition management. The graphs in [8] show that good
with and without an effective equalization process. With cells show balance to about 10 mV, while cells that deviate
equalization in action (Test pack 1), the cell-to-cell variation from this lose capacity and must be changed out.
is held to about 12 mV. Without equalization, the variation Most practical rechargeable batteries have a monotonic
begins to drift up after about 13 cycles. Fig. 3 shows the effect relationship between voltage and SOC. The characteristic in
on performance. The string capacity begins to drop as the cell- Fig. 1 is useful because it is linear and unambiguous. In other
to-cell variation rises above about 15 mV in cycle 15. A more cases, the result is more complicated. Lithium-ion cells, for
complete analysis is provided in [5]. example, show significant voltage change over the 0% to
100% SOC range. The relationship is monotonic but
nonlinear. It would be expected that equalization would be
30 beneficial for Li-ion strings. This has been confirmed in the
Average per Cell Voltage Gradient (mV)

literature [6,7]. Nickel chemistries show much flatter voltage-

25
SOC profiles, and require very precise voltage matching to
achieve good results.
Test pack 1
20 Test pack 2 III. WHAT ARE THE BENEFITS OF EQUALIZATION?

15 Modern rechargeable battery cells of all chemistries are

specified by the manufacturer for hundreds or even thousands
10
of cycles. Series strings of cells in general do not perform up
to this level. The only difference between a single-cell and
series-cell applications is management of individual cell
5
voltages. This is the basis for many assertions in the literature
that charge imbalance is the primary failure mechanism in
0 batteries [1,2,8]. In principle, a perfect equalization process
0 5 10 15 20 25
Cycle Number would ensure that a series string performs just like a single
Fig. 2. Cell voltage gradient for comparison test (from [5]). cell over time.
INTELEC 2002 Paper 32.1 3

The need for equalization of VRLA batteries is clarified in

other ways in [5]. For example in one test, several strings of
12 lead-acid batteries were cycled without equalization. The
cells were rated for at least 400 cycles, but instead the strings
provided only 25 to 30 cycles before reaching end of life. The
charge profile did not provide the “overcharge” time (which
largely involves equalization time) specified by the
manufacturer.
The conventional equalization method is to provide a
“forced overcharge” interval after the main charge sequence.
The objective is to deliver full charge into the lowest cells.
The process can be termed “passive equalization,” since it
relies on the properties of the battery cells to restore matching.
Unfortunately, passive equalization works at the expense of
gassing and dryout of the highest cells. In addition, it is a slow
process. The forced overcharge equalization process is
routinely used with lead-acid batteries. Fig. 4. 72 V battery string, charged under passive equalization (from [9]).
When this process is used, 6 V or 12 V monoblocks
become feasible. It is still true, however, that long-term
monoblock failures usually involve a single cell that has
weakened over time. For 48 V batteries, the results in [3]
imply that intervals of several weeks will be required for
equalization. Even if long intervals are available, the
overcharge exposure can lead to thermal runaway. At higher
voltage levels, results from [5] suggest that equalization time
increases as the square of the number of cells.
A more effective equalization process is needed. Even for
monoblock pairs at 24 V, an external active process to
supplement passive equalization can accelerate matching and
maintain cycle life. Figs. 4 – 7 (from [9]) compare passive and
active equalization for a 72 V battery pack. Fig. 4 shows the
voltages of the six 12 V monoblocks during charging. The
passive equalization process is slow, and the three-day
interval here shows no clear pattern. Active equalization,
shown in Fig. 5, brings the voltages together rapidly. The Fig. 5. 72 V battery string, charged under active equalization (from [9]).
operating results are emphasized in Figs. 6 and 7, which show
the standard deviation of voltages among the six monoblocks
in each string, corresponding to Figs. 4 and 5, respectively.
Notice that passive equalization is really not reducing the
voltage standard deviation, although perhaps it is starting to
fall after about 60 h.

Fig. 6. Standard deviation of cell voltages, passive equalization (from [9])

INTELEC 2002 Paper 32.1 4

equalization, there is a process to drive the cells together, so

tight initial matching is not needed. What if a failure does
occur, perhaps because of a battery defect or other problem?
Active equalization would support changeout of the defective
battery or cell without introducing extra cell mismatch. The
combination of long cycle life with the ability to change
individual units rather than a whole string has potential for
significant cost reduction.
In summary, effective equalization should
-- Extend cycle life of a series string up to that of an
individual battery or cell.
-- Avoid failure modes based on cell imbalance.
-- Permit changeout of individual cells or monoblocks
when a failure does occur.
-- Prevent weakening of string performance caused by
individual undercharged cells.
All of these provide significant cost savings in battery
Fig. 7. Standard deviation of cell voltages, active equalization (from [9]).
installations. Longer-term, it should be possible to alter the
charge process to take advantage of active equalization and
The process presented in [9] has several key advantages further extend battery cycle life.
over competing passive and active equalization technologies:
1. It is simple and direct, relying on a capacitor switching IV. WILL EQUALIZATION REALLY EXTEND BATTERY LIFE?
approach to equalization.
2. The voltage match is exact, regardless of errors or The above results suggest that cycle life can be extended at
tolerances associated with real components. least up to the level promised by manufacturers for single cells
3. The process consumes minimal energy, and can be used or monoblocks. But can this be proven with real data? In
continuously throughout charge and discharge sequences. fact, at least four published tests independently confirm the
4. The hardware is modular, and can be configured on a performance benefits of active equalization. Each is discussed
cell-by-cell or monoblock-by-monoblock basis. The individually here.
technology lends itself very well to miniaturization. In [7], a distributed charging system was used. There were
5. The inherent cost is low because no high-tolerance multiple chargers, and the effect is independent charging of
components, controls, or specialty parts are needed. small groups of cells. This is equivalent to an active
6. Equalization proceeds independent of the charge process, equalization process on monoblocks. The chemistry was
so external current and voltage limits can be set and enforced lithium-ion. Performance improvement was evident on the
without complications. very next cycle after the distributed charging system was
Assuming that equalization of voltage supports SOC added: a 2% increase in total capacity was seen after just one
matching, what can be gained? First, it would be expected that cycle. Subsequent cycles showed additional improvement.
a series string of equalized cells would show life performance In [5], active equalizers were tested with conventional
like that of an individual cell. This is potentially the most flooded lead-acid batteries. Active equalization maintained
significant result – the 28-cycle performance of the cell-to-cell matching of better than 10 mV throughout an
unequalized charge test in [5] should extend right up to the intensive one-week accelerated test – even though a low float
manufacturer’s rating of 400 cycles with proper charge limits. limit of 2.30 V/cell was used. For a second test pack, a higher
Second, failure modes associated with imbalance (repeated voltage of 2.45 V/cell was used to drive a passive equalization
undercharge of weak cells, ultimately leading to failure) are process, but the cell deviation began to rise above 10 mV after
avoided. Third, there is no need for forced overcharge as part only about six cycles. When water loss was measured over the
of a cycle (given cell-by-cell equalization). This last is test interval, the active equalization approach showed 40%
especially interesting, since it implies that perhaps the charge less water loss than the passive method. The comparative
voltage limits can be decreased when active equalization is in water loss indicates cycle life extension on the order of 66%.
place. Lower voltages make thermal runaway less likely, gas In [10], several different equalization methods were
the cells less, and should avoid the stress on strong cells compared for effects on cycle life. Some of the methods
inherent in the passive equalization process. actually degraded pack performance. A low-cost system built
Another important benefit is interchangeability. For with the technology of [9] gave a 15% cycle life improvement
example, many users recommend that series strings be built – even though the design was probably undersized for the
with tightly-matched cells. The basis for this is to start with as batteries being tested. A summary result from [10] is given in
close a match as possible, perhaps allowing more cycles Fig. 8. The “Control C” curve represents a control battery
before cell imbalance becomes severe. With active pack that uses a manufacturer-recommended charge profile.
INTELEC 2002 Paper 32.1 5

The “BMSC” curve is for the technology of [9], with exactly the test proceeds. Battery pack 1 – the only one for which the
the same charge profile. (The profile target voltage was not standard deviation is held is about 10 mV – shows far better
reduced to take advantage of active equalization.) cycle life than the others. The figure confirms that
performance degrades unless voltage imbalance is held to a
low level, and provides a linkage between voltage-based
equalization and cycle life performance. The results also
confirm that voltage balance must be better than 10 mV/cell to
provide benefits.

Fig. 8. Capacity vs. cycle for comparative equalizer test (from [10]).
In [11], cycle tests were performed on 48 V battery packs,
with and without active equalizers. Fig. 9 (from [11]) shows
the results – cycle life extension by about a factor of three. In
this case, the active equalizers are sized appropriately, and can (a) Capacity vs. cycle number for four 24 V test packs
deliver sufficient charge to maintain cell balance. The
technology used in [11] shares many performance
characteristics of [9], although it requires precise control and
is inherently more expensive. It is important to recognize that
the result in Fig. 9 confirms that a series string should be able
to reach cycle life levels similar to those of single monoblocks
with an active equalization process in place. Pack 2, with
active equalization, achieves at least 400 cycles – consistent
with the manufacturer’s rating for a single monoblock. Pack 1,
which uses only a conventional charge cycle with passive
equalization, reaches only about 140 cycles. The overall result
is a tripling of cycle life for this 48 V pack.

3DFN
ZLWKEDWWHU\HTXDOL]HUV

(b) Cell gradient (mV/cell) for four 24 V test packs

Fig. 10. Tests of four 24 V battery packs (from [12]).

(QGRI/LIH Can active equalization technology provide precise enough
N:K2XW

3DFN
 ZLWKRXWEDWWHU\HTXDOL]HUV balance to realize the benefits of equalization? Fig. 11
compares two packs, now being cycle-tested in the laboratory.

The first uses passive equalization, while the second is

supplemented with switched-capacitor active equalizers [13].
 The standard deviation is only a few millivolts. This match
has been maintained so far over several cycles, and results are

          building quickly. The technology of [9,13] matches voltages
'ULYH&\FOH exactly. Provided sufficient charge exchange is supported,
Fig. 9. Capacity vs. cycle, comparing passive and active equalization (from voltage differences well below 10 mV can be obtained.
[11]).
In applications [14,15], the benefits of equalization have
Fig. 10 (from [12]), is a telling result. In this case, four been immediately apparent. Overall pack performance
different 24 V VRLA packs have been tested with two improves noticeably as soon as equalizers are in place. In [15],
different charge profiles and various equalization strategies. a set of batteries revived from 0% SOC without developing
Fig. 10a shows the capacity as it changes over repeated cycle any imbalance, thanks to equalization.
tests, while Fig. 10b shows the voltage standard deviation as
INTELEC 2002 Paper 32.1 6

Fig. 11. Ongoing cycle test. Pack 1 uses passive equalization while Pack 2 uses active equalization.

V. CONTINUING RESULTS actively equalized pack, on the other hand, shows very tight
voltage matching that is not degrading over time. Additional
At present, long-term cycle testing continues in our data will be available at the conference.
laboratory. Fig. 12 shows an example of voltages over several
months of cycle testing. Fig. 12a are the cell voltages of the VI. SUMMARY
passively equalized pack and Fig 12b are the cell voltages for
the actively equalized pack. Missing points reflect Effective use of series strings of battery cells requires cell-
datalogging problems. In general, the cycles have been by-cell SOC matching to maintain performance. In general,
maintained without interruption for months. The gap at SOC matching can be assured through precise voltage
approximately 700 hr shows a change in charge-discharge matching, although the matching must be excellent (10 mV in
sequence to match the recommendation of the manufacturer. the case of lead-acid cells) for success. Conventional passive
In Fig. 13, a few cycles are shown in detail. The active equalization works for short series strings (six cells, and
equalization process, (Pack 2) shown in Fig 13b, brings perhaps up to twelve), but puts stress on strong cells and loses
voltages together quickly during the charge sequence and effectiveness rapidly as the series string becomes longer. If
helps keep them close even during discharge. In Pack 2, one equalization can be assured, it provides substantial benefits
of the cells was weak from the beginning. While the such as longer cycle life, fewer failure modes, and simpler
equalizers have prevented it from getting worse, they do not maintenance. These translate into major cost savings for large-
provide a “repair” function, but the cell can be changed out scale rechargeable battery applications. The technology of [9,
without causing difficulty. 13] provides perfect voltage matching without any sensing or
Fig. 14 shows an important summary result. Here, the cell control, and is the lowest-cost active equalization method
voltage standard deviation is shown at the end of each charge known. All results to date indicate that it can successfully
process. The charge sequence recommended by the deliver on the promises of performance improvements under
manufacturer (which begins at cycle 49) appears to be holding equalization.
the voltage on the passive equalized pack reasonably well,
although a slow increase appears to be underway. The
INTELEC 2002 Paper 32.1 7

Battery Pack 1 Cell Voltages

2.6

2.5

2.4

2.3

2.2

2.1
Cell voltage

1.9

1.8

1.7

1.6

1.5

1.4

1.3

1.2
0 500 1000 1500 2000 2500 3000
Hours
Fig. 12a: Passive equalization
Battery Pack 2 Cell Voltages
2.6

2.5

2.4

2.3

2.2

2.1
Cell voltage

1.9

1.8

1.7

1.6

1.5

1.4

1.3

1.2
0 500 1000 1500 2000 2500 3000
Hours
Fig. 12b: Active equalization
INTELEC 2002 Paper 32.1 8

2.5
REFERENCES
2.4
[1] Ron D. Brost, “Performance of valve-regulated lead acid batteries in
2.3 EV1 extended strings,” in Proc. IEEE Ann. Battery Conf. Appl.
Advances, 1998, pp. 25-29.
2.2
[2] J. B. Olson, E. D. Sexton, “Operation of lead-acid batteries for HEV
2.1 applications,” in Proc. IEEE Ann. Battery Conf. Appl. Advances, 2000,
Cell voltage

pp. 205-210.
2
[3] A. Lohner, E. Karden, R. W. DeDoncker, “Charge equalizing and
1.9 lifetime increasing with a new charging method for VRLA batteries,” in
Proc. IEEE Int’l. Telecommunications Energy Conf., 1997, pp. 407-411.
1.8
[4] R. O. Hammel, A. J. Salkind, D. Linden, “Sealed lead-acid batteries,” in
D. Linden, ed., Handbook of Batteries, 2nd ed. New York: McGraw-
1.7
Hill, 1995, p. 25.10.
1.6
[5] S. West, P. T. Krein, “Equalization of valve-regulated lead-acid
batteries: issues and life tests,” in Proc. IEEE Int’l. Tele-
1.5 communications Energy Conf., 2000, pp. 439-446.
2095 2100 2105 2110 2115 2120 2125 2130
[6] M. J. Isaacson, R. P. Hollandsworth, P. J. Giampaoli, F. A. Linkowsky,
Hours
A. Salim, V. L. Teofilo, “Advanced lithium ion battery charger,” in
Fig 13a: Passive equalization Proc. IEEE Ann. Batt Conf. Appl. Advances, 2000, pp. 193-198.
2.5
[7] T. T. Sack, J. C. Tice, R. Reynolds, “Segmented battery charger for high
energy 28 V lithium ion battery” in Proc. IEEE Ann. Batt Conf. Appl.
2.4
Advances, 2001, pp. 157-159.
2.3 [8] T. L. Churchill, “Battery condition management: an important way to
protect a critical asset,” in Proc. IEEE Ann. Batt Conf. Appl. Advances,
2.2 2000, pp. 71-76.

2.1 [9] C. Pascual, P. T. Krein, “Switched capacitor system for automatic series
Cell voltage

battery equalization,” in Proc. IEEE Applied Power Electronics Conf.,

2 1997, pp. 848-854.

1.9 [10] R. E. MacDougall, J. D. Bertolino, K. L. Rodden, E. T. Alger, “Lab

testing of battery charge management systems for electric and hybrid
1.8 vehicle battery packs to evaluate cycle life improvement,” in Proc. IEEE
Ann. Batt Conf. Appl. Advances, 2000, pp. 237-242.
1.7
[11] N. H. Kutkut, “Life cycle testing of series battery strings with individual
battery equalizers,” white paper, Power Designers, Inc., Madison, WI,
1.6
2000. Available: http://www.powerdesigners.com/products
/PowerCheq/HEV_Paper.pdf.
1.5

2095 2100 2105 2110 2115 2120 2125 2130 [12] S. West, “Performance evaluation of switched capacitor battery
Hours equalizer,” M.S. Thesis, University of Illinois, May 2000, p. 84.

Fig 13b: Active equalization [13] C. Pascual, P. T. Krein, “Switched capacitor systems for automatic series
15 battery equalization,”
Pack1
Pack2
U. S. Patent 5,710,504, January 1998.
12.5
[14] J. A. Cellarius, S. P. West, P. T. Krein, R. A. White, “Design and
development of a power tracking series-type hybrid vehicle,” in 1999
10 FutureCar Challenge (Warrendale, PA: Society of Automotive
Engineers, 2000).
σ (mV)

7.5 [15] J. A. Locker, private communication with respect to operating results of

Sunrayce ’97, June 1997.
5

2.5

0
20 40 60 80 100 120 140 160 180 200
Cycle number

Fig 14: Standard deviation of cell voltages at end of charge